Breacher.ai has announced the general availability of the OSES™ AI Vishing Simulator, a self-run voice phishing simulator that allows security teams to test their people and procedures using real phone calls. The product is available now starting at $199 per month.
The simulator uses an agentic AI voice agent to call employees, phone the help desk, and leave voicemails, then hold a live conversation with whoever answers. It is designed to mimic the tactics of real attackers, who increasingly use phone calls to bypass traditional email security controls.
The launch comes amid a sharp rise in voice phishing, or vishing. According to Mandiant's M-Trends 2026 report, voice phishing accounted for 11 percent of intrusions in its 2025 investigations, making it the second most common initial infection vector, ahead of email phishing at 6 percent. CrowdStrike's 2025 Global Threat Report recorded a 442 percent rise in vishing between the first and second halves of 2024.
A phone call reaches an employee with no email gateway, link scanner, or reputation score in front of it. The verification procedure is the only control in the path, and most organizations have never tested whether it holds when someone on the line pushes back. The OSES™ AI Vishing Simulator addresses this gap by providing seven capabilities in one product: real phone calls on live phone lines, outbound calls to employees posing as IT support, an executive, or a vendor; inbound calls to the service desk requesting password resets, MFA re-enrolment, or account changes; voicemail drops that set up a callback answered live by the agent in the same persona; deepfake voice using a synthetic voice or a consented clone of a real executive; agentic AI that listens, adapts, handles objections, and decides what to say next toward the objective set for the campaign; and prompt-built pretexts so a security team can describe the caller, the reason for the call, and the objective in plain language, or paste in a vishing call someone reported.
"The phone call is now the front door, and the person who answers it is working without a filter, a gateway, or a second opinion," said Jason Thatcher, Founder and CEO of Breacher.ai. "AI agents can already hold that conversation at scale. The only way to know whether your verification procedure holds is to call it the way an adversary would. We built the simulator so any security team can do that this month, not next budget cycle."
Security teams run the simulator themselves in four steps: they write the pretext as a prompt, choose a synthetic or consented cloned voice, and launch calls against users synced from Microsoft Entra ID or Google Workspace, with no endpoint agent and no phone system changes. Every call rolls into the OSES™ Risk Score, which records how far the worst call progressed (DEPTH) and how widely calls succeeded across the organization (SPREAD), benchmarked against the customer's sector.
Safeguards are built into the product. Executive voices are cloned only with the consent of the person being cloned, confirmed by a biometric check and voice verification before anything is generated, and an administrator cannot authorize cloning someone else. Every campaign runs under signed authorization, call recording is configured to the consent rules of each jurisdiction in scope, and reporting is organizational, with no named individuals.
Organizations that want email, SMS, QR, Microsoft Teams, and live video avatars orchestrated alongside voice can move to the full OSES™ Platform and keep their OSES™ Risk Score baseline. Breacher.ai also offers managed delivery for teams that want the simulator run for them.
